Having built an interpretive language for a very specialized application, I found that many times I needed to implement a "function" in the base language. Most of the time these implementations of "functions" were to make things "easier" at the interpretive language level, some were "one-time" things that probably could have been done as some code that was interpreted. There were always trade-offs. Some were more apparent than others (doing it inside the interpreter was MUCH faster). Not knowing the mind set of $Larry, I assume that (at least in early iterations) the same thing took place, which yields the "More than one way" syndrome we all enjoy. So, to know all 200+ functions isn't the best use of ones time, but every once in a while, scanning a book might yield some insight.
It all boils down to knowing that while it probably IS possible to implement a regular expression engine in Perl [without using regular expressions], the language has that "stuff" already there, and understanding this helps considerably.